New Frameworks Advance Agent-Based Systems for RL Modeling, Epidemic Planning, and Action Recognition

Three recent arXiv papers showcase advances in automated AI systems across diverse applications.

A-LAMP: Automated RL Framework

According to arXiv paper 2512.11270v1, researchers have developed A-LAMP, an “Agentic LLM-Based Framework for Automated MDP Modeling and Policy Generation.” The paper addresses a key challenge in applying reinforcement learning to real-world tasks: converting informal descriptions into formal Markov decision processes (MDPs), implementing executable environments, and training policy agents. The abstract notes that “automating this process is challenging,” though the paper does not detail specific results or methodology beyond this framing.

EpiPlanAgent for Epidemic Response

A separate paper (arXiv:2512.10313v2) introduces EpiPlanAgent, described as “an agent-based system using large language models (LLMs) to automate the generation” of epidemic response plans. According to the abstract, the research aims to address the traditionally “labor-intensive manual methods” used in epidemic response planning.

Skeleton-Cache for Action Recognition

The third paper (arXiv:2512.11458v1) presents Skeleton-Cache, described as “the first training-free test-time adaptation framework for skeleton-based zero-shot action recognition (SZAR).” According to the researchers, the framework aims to improve “model generalization to unseen actions during inference.”

All three papers represent works in progress, with full methodologies and results available in their respective preprints.